Visualizing Manuscript connections with Palladio

Geographical dispersion of autograph manuscripts of Ibn Ṭūlūn

How do you picture the work in the archives? Dusty pages, creaking in the half-light of old halls of knowledge? Bent-over people shuffling their feet between rows of shelves? The creaking of a microfilm viewer seeming deafening in a silence otherwise interrupted only by occasional coughs? It is probably all happening somewhere and might be a strong impetus behind advocation for the digitization of primary sources.

Yet, the work in the archives is also one of the most intriguing parts of an historian’s work. There you encounter the handwriting of a scholar whose ideas you have followed for so long, and the manuscript closes the gap in time, occasionally bearing traces of its own travels and trajectories through time and space (the result looks a lot like space as well, with stars and planets).

One interesting aspect of Ibn Ṭūlūn’s autograph corpus is that most of what survives are rather small-scale writings. In several cases, these are found within one volume entirely consisting of his own works. Moreover, between these volumes there are occasionally several autographs of the same work. And these – both textual and material items – are today dispersed in libraries in Egypt, Lebanon, Syria, Turkey, Germany, the Netherlands, England, Ireland, and the United States (more copies in other hands are found elsewhere). In order to make sense of the connections between manuscripts, on one hand, and texts, on the other, as well as their locations today, I have used Palladio.

The present post tries to convey this excitement through the use of visualizations of the dispersion of Ibn Ṭūlūn’s works and codices . I have used Palladio, a network visualization software created at Stanford University. It is platform independent and runs in your browser. It is less powerful than other network analysis and visualization software but it allows you to get results in your first day, which can be very motivational. There is also a good tutorial by Marten Düring, which leads you through the first steps. For the first steps of setting up a Palladio project, you should refer to those sites.

My own project consists of three csv sheets (the full json project file cannot be uploaded here):

Manuscripts Ibn Tulun-Tabelle 1 Works Ibn Tulun-Tabelle 1 Relations-Tabelle 1

Follow Fig. 1 (left to right) in uploading them. The figure also shows the connections (“extension”) you should create between them: Manuscripts should be connected to call number, and key should connect in tables Works and Relations.

Fig. 1: Palladio project start screen

The following visualizations have been created using the graph function. The first image shows the collections which hold Ibn Ṭūlūn autographs and which hold overlapping works. The figure was created by using Place as Source and Place_2 as Target.

Fig. 2: Connections between collections, created through holding of identical works.

The Dār al-Kutub, Cairo, and the Chester Beatty Library, Dublin, clearly emerge as the main collections for Ibn Ṭūlūn autographs, with Alexandria, Leiden, and the Staatsbibliothek Berlin trailing.

Fig. 3: Manuscripts held in collections and overlaps in contained works.

Figure 3 makes the interconnections between the holdings of the two main collections even more obvious. Among the manuscripts held by either are several works that can be found in manuscripts of the other as well. The Princeton cluster, which now appears more prominent, contains several fragments or other drafts of texts, held at either Dublin or Cairo, as Kristina Richardson has recently proved. The British Library, London, is connected to this through their one partial autograph, whose remaining volumes are held in Cairo. The size of the Leiden cluster derives from the fact that each text is counted as one manuscript, even though some only count 5 folios.

Fig. 4: Fig. 3: Works held in collections and overlaps through appearing in other manuscripts.

The centrality of the Cairo connection becomes yet clearer if we switch the focus from manuscript codices to individual works. Dār al-Kutub holds the largest number of Ibn Ṭūlūn’s multiple-text manuscripts (MTM), which each contain up to more than ten titles. (I have used reference keys here, since titles are too long to display and Palladio does not allow for proper transliteration. But I am not talking about content here, anyway.) For the same reason, the Alexandria collection (holding only one MTM) draws almost even with the one in Leiden.

Also, the overlap between the Cairo and Dublin collections is now so strong that reading anything becomes difficult (also within the Cairo cluster). Therefore, I have included one last figure. This shifts the focus from the collections to the direct relations between works and manuscripts. In this case, I have used Palladio’s facet filters to consider only those manuscripts held in Cairo, Dublin, and Princeton. Figure 5 is, however, only a limited screenshot which captures the most interesting cases of overlap between – and within – these collections.

Fig. 5: Manuscripts held at Cairo, Dublin, Princeton, containing identical textual items (dark dots are manuscripts, light dots works).

Three out of four MTMs held in Dublin (MS CBL…) have strong overlaps with Cairene manuscripts (MTMs from the Taymur Pasha collection, MS TayMa…). And also within the Taymur collection, the MTMs 759 and 373 share 11 identical items  between them.


Digital possibilities for Research on Arabic Literature

The word ‘digital humanities’ is ringing in all our ears these days. And indeed, this umbrella term encompasses a number of original – and sometimes breathtaking – approaches to old corpora. Yet, reality often enough stays far behind expectations, and the Arabic language and script, in particular, cause obstacles for applications in our field.

Thus, many institutions are happy to upload images of manuscripts to a repository and call it a day, often even without providing much needed meta data to situate the item at hand within time and space. One of the most insightful and engaging series on the risks and downsides of this development remains Dagmar Riedel’s blog Islamic Books.

Now, the UK based kitab project promises a remedy to this sorry state. The premises of the project are indeed promising. A large – and growing – corpus that apparently consists of source texts that have been taken from shamela and enhanced for scholarly work are made available for searches for text passages and how they might have been reused by other authors. The project promises that their tools will allow to “discover relationships between these texts and also the profoundly intertextual circulatory systems in which they sit.”

This would be a great step forward. The intertextuality of Arabic literature, and writing in general, has long been acknowledged, but due to the magnitude of the corpora has so far obstructed an overall assessment of the importance of certain authors on later generations.

Until now, the project is restricted to closed community of users, a beta version so to speak. Unfortunately, the website offers no detailed information on how the available corpus is composed. Whereas I am confident that the ‘classical period’ is well-represented, I wonder in how far that applies to the Middle Periods, especially towards the end of the period covered (750-1500).

The project organizers can certainly not be blamed that large parts of the literature of the later centuries have not even been edited yet. On the contrary, they even urge scholars to add new texts to the corpus. But the risk remains that such a handy tool as the kitab project offers, might reinforce earlier survival, edition, and digitization biases. The already vast amount of 4,500 individual texts it allows to access and analyze easily, might be too tempting as to consult other works which are only available in some obscure collection overseas.

It should be our shared interest as researchers that this issue is addressed and that the kitab project (as well as other DH projects) rather lead to the opposite direction. That they proliferate the research on unedited manuscripts by easing the access to those sources, which have been canonized through different stages of scholarly engagement (again, the editorial and digitizing practices play a role here). In sharing our own work within the framework of kitab might be a better way reinsert those texts into the assessment of scholarly traditions than simple provision of digital images.

AMENDMENT: More detailed information on the composition of the corpus used by kitab can be found on the github page of the Leipzig based OpenArabic Project. And indeed, there is a clear bias towards the earlier centuries and the ‘classical period’ in particular. Under the thematic rubric ‘biographies’, the 14th century appears as the heyday whereas the 15th and 16th centuries seem all but void of that sort of literature (under biographical works, the number drops from 24 titles for the 14th to 7 for the 15th century). Chronicles are rather absent from the corpus in general, with the highest peak (again in the 14th century) only reaching a number of 6. For the whole 15th century, only two chronicles are included and none for the 1500s.